Public Advisory Session

AI Security Vulnerability Red Team Assessment

Our new customer-facing AI assistant was just found to be vulnerable to prompt-injection that can leak data between tenants. I need the board to run a red-team threat model of the assistant — attack surfaces, data-exfiltration paths, and the security controls + incident-response plan we need before we notify customers.

September 8, 2026 · Brainstorm · 3 rounds · 11 advisor messages